Dear Spencer, Stan, Plese find attached an attempt of Floquet analysis session file. There are some comments in it regarding the aspects I do not quite know how to implement, namely: 1) The implementation differences between using Arpack or ModifiedArnoldi 2) The use of SingleMode and MultipleMode. Shouldn't HalfMode be sufficient for the 3D stability analysis of a 2D base flow? 3) How is the base state dpecified if it consists of a collection of snapshots along a period. 4) Is an initial condition required? 5) If the base field is 2D, do I need to set w=0 for the base state (and initial boundary condition)? 6) Do I need to set up a forcing "StabilityCoupledLNS" for the Floquet analysis as in the standard linear stability analysis of steady solutions? The mesh file is quite large so I don't attach it. If you want to toy with it, I'll prepare a lighter case that can run faster. Cheers On 21/05/17 20:27, Sherwin, Spencer J wrote:
Dear Fernando, Stan,
Thanks for your query and Stan’s answers too. As you have both stated the Floquet problem is indeed for time dependent base flows that are intended to be interpolated with a Fourier approximation in time.
Unfortunately I have to had anyone working on this aspect of the code for a while. It is clear now we should set up a demo with a Floquet example to help answer this although I do not know we will get this done very soon due to pressure on other developments.
If you have an example file we could perhaps have a look at your test case to see if we can help and then maybe set this up as a tutorial or test case.
Cheers, Spencer.
On 18 May 2017, at 12:00, F Mellibovsky <fernando.mellibovsky@upc.edu <mailto:fernando.mellibovsky@upc.edu>> wrote:
Thanks Stan,
My 2D solution is indeed time-periodic and stable to all spanwise-independent perturbations (at the Reynolds considered) . It is analogous to the vortex-shedding cylinder case, which becomes tridimensional through the destabilisation of the so-called mode A (a real eigenvalue that crosses the unit circle at +1 and breaks the spanwise invariance of the solution). The setup for the example that appears in the user guide would come very handy because that's exactly what I need to do.
Cheers
On 18/05/17 12:09, Stanisław Gepner wrote:
Hi Fernando,
Not a direct answer to your question, as I have not used Floquet, sorry.
From your description I think I was doing sth. similar, with a stationary flow periodic in space and unstable to stream- and spanwise perturbations. I ended up tracking the growth of the modal energy, like you. The Floquet, as implemented in Nektar is, I think, more for time dependent problems. (Can one of the Devs. comment?).
As for the Arpack and ModifiedArnoldi Drivers, my observations are such:
1. MA works in parallel, converges a bit slower than Arpack, does not use shifting (?), if kdim is to large it might produce physical values (not sure why, I always thought that a larger Krylov space is better).
2. Arpack converges faster, but sometimes the unstable mode is "hidden" deep in the spectrum, so one needs to retrieve a lot of EVs and find the correct one. You can use shifting, if you know an approximation of the complex growth rate, to move the interesting mode "forward".
3. Arpack does not work in parallel, you will get a solution but it will make no sense.
Best,
Stan Gepner
On 05/18/2017 11:33 AM, F Mellibovsky wrote:
Dear all,
I have computed a stable *2D periodic solution* with nektar++ time evolving the *incompressible Navier-Stokes solver*. Now I would like to analyse it's *stability to spanwise-dependent perturbations*. I have started doing this with time evolution, by using a homogeneous spanwise direction with just one mode and monitoring its growth rate. This is however extremely slow. I have noticed that in the user guide there's an example that does Floquet stability analysis on the periodic 2D vortex shedding solution of the flow around a cylinder, but unfortunately I haven't manage to find the conditions xml file used for the computations. Does anyone have it, so that I can use it to set up my *Floquet analysis*? I have doubts as to how a sequence of snapshots along a period are loaded and also about the differences between using Arpack or modified Arnoldi, among other issues. I would be very greatful if someone can send me an conditions file exemplifying the implementation of Floquet analysis. Many thanks in advance.
Cheers,
Fernando Mellibovsky
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk <mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
Spencer Sherwin McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ
s.sherwin@imperial.ac.uk <mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052